feat: 应用名称更新为灵可析并增强 Chat 功能

- 更新 Android/iOS 应用名称和图标为灵可析
- Chat 支持取消正在运行的 Agent 对话
- 改进 ChatBloc 状态管理（区分发送/等待/流式/取消状态）
- HomeScreen 支持外部注入 ChatBloc 和显示等待指示器
- 后端 Agent 运行服务优化（消息处理、usage 追踪）
- 补充相关单元测试和 Widget 测试

backend/tests/unit/core/agent/test_litellm_usage.py

@@ -5,15 +5,14 @@ import pytest
 from core.agent.infrastructure.litellm.usage_tracker import extract_usage_and_cost
 
 
-def test_usage_tracker_extracts_tokens_and_cost(
-    monkeypatch: pytest.MonkeyPatch,
-) -> None:
-    monkeypatch.setattr(
-        "core.agent.infrastructure.litellm.usage_tracker.completion_cost",
-        lambda completion_response: 0.123,
-    )
+def test_usage_tracker_uses_custom_pricing_for_qwen35() -> None:
     response = {
-        "usage": {"prompt_tokens": 11, "completion_tokens": 7, "total_tokens": 18},
+        "model": "dashscope/qwen3.5-flash",
+        "usage": {
+            "prompt_tokens": 11,
+            "completion_tokens": 7,
+            "total_tokens": 18,
+        },
     }
 
     usage = extract_usage_and_cost(response)
@@ -21,7 +20,8 @@ def test_usage_tracker_extracts_tokens_and_cost(
     assert usage.prompt_tokens == 11
     assert usage.completion_tokens == 7
     assert usage.total_tokens == 18
-    assert usage.cost == 0.123
+    assert usage.cost == pytest.approx(0.0000162)
+    assert usage.cost_source == "custom_pricing"
 
 
 @pytest.mark.parametrize(
@@ -33,19 +33,10 @@ def test_usage_tracker_extracts_tokens_and_cost(
     ],
 )
 def test_usage_tracker_falls_back_to_local_qwen35_pricing_when_model_unmapped(
-    monkeypatch: pytest.MonkeyPatch,
     prompt_tokens: int,
     completion_tokens: int,
     expected_cost: float,
 ) -> None:
-    def _raise_unmapped(*, completion_response):  # type: ignore[no-untyped-def]
-        del completion_response
-        raise Exception("This model isn't mapped yet")
-
-    monkeypatch.setattr(
-        "core.agent.infrastructure.litellm.usage_tracker.completion_cost",
-        _raise_unmapped,
-    )
     response = {
         "model": "dashscope/qwen3.5-flash",
         "usage": {
@@ -59,3 +50,22 @@ def test_usage_tracker_falls_back_to_local_qwen35_pricing_when_model_unmapped(
 
     assert usage.cost == pytest.approx(expected_cost)
     assert usage.cost_source == "custom_pricing"
+
+
+def test_usage_tracker_uses_cached_pricing_for_deepseek_chat() -> None:
+    response = {
+        "model": "deepseek/deepseek-chat",
+        "usage": {
+            "prompt_tokens": 1_000_000,
+            "completion_tokens": 100_000,
+            "total_tokens": 1_100_000,
+            "prompt_tokens_details": {
+                "cached_tokens": 400_000,
+            },
+        },
+    }
+
+    usage = extract_usage_and_cost(response)
+
+    assert usage.cost == pytest.approx(1.58)
+    assert usage.cost_source == "custom_pricing"

backend/tests/unit/core/agent/test_run_resume_service.py

@@ -1058,6 +1058,128 @@ async def test_run_service_executes_backend_calendar_tool_and_emits_result(
     assert runtime_state["status"] == AgentChatSessionStatus.COMPLETED
 
 
+@pytest.mark.asyncio
+async def test_run_service_does_not_persist_model_code_for_user_message(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    session_id = uuid4()
+    user_id = uuid4()
+    captured: dict[str, object] = {}
+    message_calls: list[dict[str, object]] = []
+
+    class _FakeDbSession:
+        async def commit(self) -> None:
+            return None
+
+    class _FakeSessionFactory:
+        def __call__(self) -> "_FakeSessionFactory":
+            return self
+
+        async def __aenter__(self) -> _FakeDbSession:
+            return _FakeDbSession()
+
+        async def __aexit__(self, exc_type, exc, tb) -> bool:
+            del exc_type, exc, tb
+            return False
+
+    class _FakeSessionRepository:
+        def __init__(self, session: object) -> None:
+            del session
+
+        async def lock_session_for_update(self, *, session_id: object):
+            return SimpleNamespace(
+                id=session_id,
+                user_id=user_id,
+                status=AgentChatSessionStatus.PENDING,
+                message_count=0,
+                total_tokens=0,
+                total_cost=0,
+                state_snapshot=None,
+            )
+
+        async def next_message_seq(self, *, session_id: object):
+            del session_id
+            return 1
+
+        async def update_runtime_state(self, **kwargs) -> None:
+            captured["update_runtime_state"] = kwargs
+
+    class _FakeMessageRepository:
+        def __init__(self, session: object) -> None:
+            del session
+
+        async def append_message(self, **kwargs) -> None:
+            message_calls.append(kwargs)
+
+    class _FakeRuntime:
+        def execute(
+            self,
+            *,
+            user_input: str,
+            system_prompt: str | None = None,
+            tools: list[dict[str, object]] | None = None,
+        ):
+            del user_input, system_prompt, tools
+            return {
+                "assistant_text": "ok",
+                "prompt_tokens": 1,
+                "completion_tokens": 1,
+                "total_tokens": 2,
+                "cost": "0.001",
+                "agui_events": [],
+            }
+
+    async def _fake_load_agent_model_selection(self, _session):
+        del self
+        return ("qwen3.5-flash", "dashscope", SystemAgentLLMConfig())
+
+    async def _fake_load_user_agent_context(self, session, session_id, user_id):
+        del self, session, session_id
+        return SimpleNamespace(
+            user_id=user_id,
+            username="demo-user",
+            bio=None,
+            settings=SimpleNamespace(
+                preferences=SimpleNamespace(
+                    interface_language="zh-CN",
+                    ai_language="zh-CN",
+                    timezone="Asia/Shanghai",
+                    country="CN",
+                )
+            ),
+        )
+
+    monkeypatch.setattr(
+        "core.agent.application.run_service.SessionRepository",
+        _FakeSessionRepository,
+    )
+    monkeypatch.setattr(
+        "core.agent.application.run_service.MessageRepository",
+        _FakeMessageRepository,
+    )
+    monkeypatch.setattr(
+        "core.agent.application.run_service.create_runtime",
+        lambda **_kwargs: _FakeRuntime(),
+    )
+    monkeypatch.setattr(
+        "core.agent.application.run_service.RunService._load_agent_model_selection",
+        _fake_load_agent_model_selection,
+    )
+    monkeypatch.setattr(
+        "core.agent.application.run_service.RunService._load_user_agent_context",
+        _fake_load_user_agent_context,
+    )
+
+    service = RunService(session_factory=_FakeSessionFactory())  # type: ignore[arg-type]
+    await service.run(
+        run_input=_build_run_input(thread_id=str(session_id), text="hello")
+    )
+
+    user_message = message_calls[0]
+    assert user_message["role"] == AgentChatMessageRole.USER
+    assert "model_code" not in user_message
+
+
 @pytest.mark.asyncio
 async def test_load_user_agent_context_parses_profile_settings_v1() -> None:
     session_id = uuid4()

backend/tests/unit/core/agent/test_runtime_stage_runner_usage.py

@@ -0,0 +1,72 @@
+from __future__ import annotations
+
+from types import SimpleNamespace
+
+import pytest
+
+from core.agent.infrastructure.crewai.runtime_stage_runner import (
+    LiteLLMUsageCaptureCallback,
+    extract_usage_from_captured_payload,
+    extract_usage_from_crew_output,
+)
+
+
+def test_extract_usage_from_crew_output_uses_custom_deepseek_pricing() -> None:
+    output = SimpleNamespace(
+        token_usage=SimpleNamespace(
+            prompt_tokens=1_000_000,
+            completion_tokens=100_000,
+            total_tokens=1_100_000,
+            cached_prompt_tokens=400_000,
+        )
+    )
+
+    usage = extract_usage_from_crew_output(
+        output=output,
+        model="deepseek/deepseek-chat",
+    )
+
+    assert usage.prompt_tokens == 1_000_000
+    assert usage.completion_tokens == 100_000
+    assert usage.total_tokens == 1_100_000
+    assert usage.cost == pytest.approx(1.58)
+
+
+def test_extract_usage_from_captured_payload_uses_custom_pricing() -> None:
+    usage = extract_usage_from_captured_payload(
+        captured_usage={
+            "prompt_tokens": 1_000_000,
+            "completion_tokens": 100_000,
+            "total_tokens": 1_100_000,
+            "prompt_tokens_details": {"cached_tokens": 400_000},
+        },
+        model="deepseek/deepseek-chat",
+    )
+
+    assert usage.prompt_tokens == 1_000_000
+    assert usage.completion_tokens == 100_000
+    assert usage.total_tokens == 1_100_000
+    assert usage.cost == pytest.approx(1.58)
+
+
+def test_usage_capture_callback_extracts_nested_usage_payload() -> None:
+    callback = LiteLLMUsageCaptureCallback()
+
+    callback.log_success_event(
+        kwargs={},
+        response_obj={
+            "usage": {
+                "prompt_tokens": 15,
+                "completion_tokens": 9,
+                "total_tokens": 24,
+            }
+        },
+        start_time=0,
+        end_time=0,
+    )
+
+    assert callback.captured_usage == {
+        "prompt_tokens": 15,
+        "completion_tokens": 9,
+        "total_tokens": 24,
+    }